You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Patrick Wendell (JIRA)" <ji...@apache.org> on 2015/06/01 02:21:17 UTC
[jira] [Closed] (SPARK-7959) Uneven distribution of receivers in
the cluster
[ https://issues.apache.org/jira/browse/SPARK-7959?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Patrick Wendell closed SPARK-7959.
----------------------------------
Resolution: Not A Problem
The scheduling of receivers is not deterministic and based on Spark's underlying scheduler, so I don't think this violates existing expectations.
If the request is to add a new feature that gives some control over receiver scheduling (probably a good idea), please feel free to open a new feature JIRA.
> Uneven distribution of receivers in the cluster
> -----------------------------------------------
>
> Key: SPARK-7959
> URL: https://issues.apache.org/jira/browse/SPARK-7959
> Project: Spark
> Issue Type: Bug
> Components: Streaming
> Affects Versions: 1.3.1, 1.4.0
> Reporter: Nishkam Ravi
> Attachments: Screen Shot 2015-05-29 at 3.52.01 PM.png
>
>
> Receivers are not getting scheduled as expected. Attached is a screenshot for 8 receivers running on a cluster with 4 workers. In some cases, all receivers get scheduled on a single node. When the receivers get distributed evenly, the throughput can be upto 25% better (for the workload experimented with).
> Not sure if this is a known issue and I haven't investigated it further. But given that I do see perfectly well distributed receivers (in round robin fashion) at times, my hunch is that the receivers are racing with other jobs/tasks being scheduled.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org